AAAI 2024 Papers — Page 11
AAAI Conference on Artificial Intelligence · 2331 papers
Hyperbolic Graph Diffusion Model
Lingfeng Wen (East China Normal University), Xian Wei (East China Normal University)
GenerationData SynthesisGraph Neural NetworkDiffusion modelScore-based ModelAuto EncoderGraph
🎯 What it does: This paper proposes the Hyperbolic Graph Diffusion Model (HGDM), a two-stage graph generation method: first, it encodes nodes into hyperbolic space using a Hyperbolic Variational Autoencoder (HVAE), and then learns the joint distribution of nodes and the adjacency matrix in that space using a score-based diffusion model, thereby generating graphs with hierarchical structures.
Hypercorrelation Evolution for Video Class-Incremental Learning
Sen Liang (University of Science and Technology of China), Yang Cao (University of Science and Technology of China)
ClassificationRecognitionKnowledge DistillationConvolutional Neural NetworkVideo
🎯 What it does: A hyper-correlated evolution (HCE) framework based on hierarchical aggregation and correlation refinement is proposed for class-incremental learning in video classification.
HyperEditor: Achieving Both Authenticity and Cross-Domain Capability in Image Editing via Hypernetworks
Hai Zhang (East China Normal University), Wenming Cao (Shenzhen University)
Image TranslationGenerationDomain AdaptationGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: Utilizing hypernetworks to reallocate the weights of the StyleGAN2 generator, enabling attribute editing of real images and cross-domain style transfer.
HyperFast: Instant Classification for Tabular Data
David Bonet (Universitat Polit'cnica de Catalunya), Alexander G. Ioannidis (Stanford University)
ClassificationMeta LearningTabular
🎯 What it does: We propose HyperFast, a pre-trained hypernetwork that can generate lightweight neural network weights for new tasks based on a given support set in a single forward pass, enabling instant classification.
Hypergraph Joint Representation Learning for Hypervertices and Hyperedges via Cross Expansion
Yuguang Yan (Guangdong University of Technology), Ruichu Cai (Guangdong University of Technology)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes a Cross Expansion method that transforms hypergraphs into standard graphs and performs Graph Convolutional Network (GCN) learning on this graph, ultimately obtaining representations of hypervertices and hyperedges in the same embedding space; it also incorporates hypergraph reconstruction loss to preserve structural information.
Hypergraph Neural Architecture Search
Wei Lin (Xiamen University), Taisong Jin (Xiamen University)
ClassificationNeural Architecture SearchGraph Neural NetworkGraph
🎯 What it does: Proposed HyperNAS, a framework for automatically searching for optimal hypergraph neural network architectures;
Hypergraph-Guided Disentangled Spectrum Transformer Networks for Near-Infrared Facial Expression Recognition
Bingjun Luo (Tsinghua University), Yue Gao (Tsinghua University)
RecognitionTransformerImageMultimodality
🎯 What it does: This study investigates how to achieve facial expression recognition under near-infrared (NIR) conditions and proposes a novel network model called NFER-Former.
Hyperspectral Image Reconstruction via Combinatorial Embedding of Cross-Channel Spatio-Spectral Clues
Xingxing Yang (Hong Kong Baptist University), Zaifeng Yang (Agency for Science Technology and Research)
RestorationTransformerImage
🎯 What it does: For the reconstruction of hyperspectral images from RGB images, a CESST framework is proposed, which independently extracts the spatial-spectral features of each RGB channel in a high-dimensional embedding space, and then achieves cross-channel information fusion through a combinatorial attention module.
Hypothesis, Verification, and Induction: Grounding Large Language Models with Self-Driven Skill Learning
Shaohui Peng (Intelligent Software Research Center, Institute of Software, Chinese Academy of Sciences), Ling Li (University of Chinese Academy of Sciences)
Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningAgentic AIText
🎯 What it does: Developed the HYVIN framework, which utilizes LLM to autonomously generate sub-goals and check functions, first validating feasibility in the environment, then clustering to learn general skills, and ultimately using these skills to complete instruction-following tasks.
i-Rebalance: Personalized Vehicle Repositioning for Supply Demand Balance
Haoyang Chen (Southeast University), Yan Lyu (Nanjing University of Science and Technology)
Recommendation SystemRecurrent Neural NetworkReinforcement LearningTabularTime Series
🎯 What it does: This paper proposes i-Rebalance, a personalized vehicle repositioning method based on deep reinforcement learning, which considers drivers' cruising preferences and makes decisions based on whether drivers accept recommendations, aiming to achieve supply-demand balance and enhance driver income.
ICAR: Image-Based Complementary Auto Reasoning
Xijun Wang (University of Maryland), Shan Yang (Amazon)
RetrievalTransformerContrastive LearningImage
🎯 What it does: A visual compatibility learning framework called ICAR is proposed, which can retrieve cross-domain compatible item sets from scene images.
Identifiability of Direct Effects from Summary Causal Graphs
Simon Ferreira (ENS de Lyon), Charles K. Assaad (EasyVista)
Time Series
🎯 What it does: This paper studies how to determine and estimate the direct causal effect of one time point on another in a linear dynamic structural causal model (SCM) given only the summary causal graph (SCG) of a time series.
Identification for Tree-Shaped Structural Causal Models in Polynomial Time
Aaryan Gupta (Indian Institute of Technology Bombay), Markus Bläser (Saarland University)
RecognitionComputational Efficiency
🎯 What it does: A randomized polynomial-time algorithm is proposed to determine the identifiability of each causal parameter in tree-shaped structural causal models (tree-shaped SCMs) and provides the corresponding symbolic expressions.
Identification of Causal Structure in the Presence of Missing Data with Additive Noise Model
Jie Qiao (Guangdong University of Technology), Zhifeng Hao (Shantou University)
TabularBiomedical Data
🎯 What it does: This paper proposes a framework for causal structure learning using the Additive Noise Model (ANM) under the condition of weak self-masking.
Identification of Causal Structure with Latent Variables Based on Higher Order Cumulants
Wei Chen (Guangdong University of Technology), Kun Zhang (Carnegie Mellon University)
Tabular
🎯 What it does: A complete method is proposed to identify the causal structure (whether causal edges exist and their direction) of two observed variables under the influence of latent variables using higher-order cumulants.
Identification of Necessary Semantic Undertakers in the Causal View for Image-Text Matching
Huatian Zhang (University of Science and Technology of China), Zhendong Mao (University of Science and Technology of China)
RetrievalRecurrent Neural NetworkVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes a causal perspective-based image-text matching method, which first theoretically derives the probability of the necessity of semantic sharing degree for segment pairs (PN_f), and implements a Necessary Undertaker Identification Framework (NUIF). It filters the most critical visual/language segments for matching through adaptive representation and two quantification methods of PN_f (PN_f‑d and PN_f‑r).
iDet3D: Towards Efficient Interactive Object Detection for LiDAR Point Clouds
Dongmin Choi (Letsur Inc), Jaegul Choo (Korea Advanced Institute of Science and Technology)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: An interactive 3D object detection framework iDet3D is proposed, which allows users to quickly label multi-class objects in LiDAR point clouds by clicking on a 2D interface.
IGAMT: Privacy-Preserving Electronic Health Record Synthesization with Heterogeneity and Irregularity
Wenjie Wang (ShanghaiTech University), Li Xiong (Emory University)
Data SynthesisSafty and PrivacyTransformerGenerative Adversarial NetworkBiomedical DataElectronic Health Records
🎯 What it does: A framework named IGAMT is proposed for synthesizing electronic health record data with temporal heterogeneity, missing values, and heterogeneous features while ensuring differential privacy.
IINet: Implicit Intra-inter Information Fusion for Real-Time Stereo Matching
Ximeng Li (Huazhong University of Science and Technology), Wenbing Tao (Huazhong University of Science and Technology)
Depth EstimationComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: A real-time stereo matching framework IINet based on implicit 2D networks is proposed, which achieves high-precision and low-latency disparity estimation through internal information fusion and efficient resolution enhancement.
Image as a Language: Revisiting Scene Text Recognition via Balanced, Unified and Synchronized Vision-Language Reasoning Network
Jiajun Wei (East China Normal University), Umapada Pal (Indian Statistical Institute)
RecognitionTransformerVision Language ModelImageText
🎯 What it does: This paper proposes BUSNet, a balanced unified synchronous visual-language reasoning network that treats images as noisy text for length dimension concatenation, improving scene text recognition.
Image Captioning with Multi-Context Synthetic Data
Feipeng Ma (University of Science and Technology of China), Xiaoyan Sun (University of Science and Technology of China)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelImageTextChain-of-Thought
🎯 What it does: This paper proposes a complete process for generating and training multi-context synthetic data, utilizing large language models (LLM) and diffusion models (Stable Diffusion) to synthesize multi-perspective descriptions and corresponding images, and trains an image captioning model solely using these synthetic image-text pairs.
Image Content Generation with Causal Reasoning
Xiaochuan Li (Inspur Electronic Information Industry Co), Rengang Li (Inspur Electronic Information Industry Co)
GenerationTransformerLarge Language ModelDiffusion modelContrastive LearningImageMultimodality
🎯 What it does: This paper proposes the Visual Question Answering and Image (VQAI) task and constructs a corresponding dataset, while also designing a latent space-guided causal image generation method (LGD) based on large language models and diffusion models.
Image Safeguarding: Reasoning with Conditional Vision Language Model and Obfuscating Unsafe Content Counterfactually
Mazal Bethany (University of Texas at San Antonio), Peyman Najafirad
SegmentationExplainability and InterpretabilityTransformerLarge Language ModelVision Language ModelImage
🎯 What it does: A Conditional Visual-Language Model (ConditionalVLM) has been developed to generate accurate justifications for unsafe images, and a sub-object segmentation algorithm based on adversarial explanations (CSE) has been proposed to minimize the masking of unsafe areas while keeping safe areas unchanged.
Imagine, Initialize, and Explore: An Effective Exploration Method in Multi-Agent Reinforcement Learning
Zeyang Liu (Xi'an Jiaotong University), Xuguang Lan (Xi'an Jiaotong University)
TransformerReinforcement LearningPrompt EngineeringSequential
🎯 What it does: This paper proposes a method called Imagine, Initialize, and Explore (IIE) that utilizes Transformers to generate key interaction states and initializes multiple agents in these states, thereby enhancing exploration efficiency in complex collaborative tasks.
Imitate the Good and Avoid the Bad: An Incremental Approach to Safe Reinforcement Learning
Huy Hoang (Singapore Management University), Pradeep Varakantham (Singapore Management University)
Safty and PrivacyReinforcement Learning
🎯 What it does: This paper proposes a safety reinforcement learning framework that does not require estimating cost constraints, using an incremental good trajectory imitation and bad trajectory avoidance method to directly improve the policy;
Imitation of Life: A Search Engine for Biologically Inspired Design
Hen Emuna (Hebrew University of Jerusalem), Dafna Shahaf (Hebrew University of Jerusalem)
RetrievalTransformerLarge Language ModelText
🎯 What it does: An automated search engine named BARCODE has been developed to mine bio-inspired solutions corresponding to engineering challenges from large-scale natural language texts such as Wikipedia.
Impartial Adversarial Distillation: Addressing Biased Data-Free Knowledge Distillation via Adaptive Constrained Optimization
Dongping Liao (University of Macau), Chengzhong Xu (University of Macau)
ClassificationKnowledge DistillationGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes an unbiased adversarial distillation method (IPAD) for the scenario where the teacher model in data-free knowledge distillation (DFKD) comes from an imbalanced dataset, addressing the issues of the generator's excessive focus on minority classes and the collapse of the main class patterns.
Implications of Distance over Redistricting Maps: Central and Outlier Maps
Seyed A. Esmaeili (University of Chicago), Brian Brubach (Columbia University)
Graph
🎯 What it does: This paper proposes an interpretable and computable distance metric for redistricting maps, and defines the Medoid and Centroid based on this metric to detect the gerrymandering phenomenon in district partitioning.
Implicit Modeling of Non-rigid Objects with Cross-Category Signals
Yuchun Liu (United Imaging Intelligence), Ziyan Wu (United Imaging Intelligence)
SegmentationGenerationData SynthesisOptimizationBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A multi-object deep implicit function MODIF is proposed for simultaneously modeling the shapes of multiple non-rigid objects (such as organs) and their interrelations.
Improve Robustness of Reinforcement Learning against Observation Perturbations via l∞ Lipschitz Policy Networks
Buqing Nie (Shanghai Jiao Tong University), Yue Gao (Shanghai Jiao Tong University)
Reinforcement Learning
🎯 What it does: Proposes the SortRL method, which utilizes l∞ Lipschitz continuous SortNet to improve the robustness of DRL under observation perturbations.
Improved Anonymous Multi-Agent Path Finding Algorithm
Zain Alabedeen Ali (Moscow Institute of Physics and Technology), Konstantin Yakovlev (Federal Research Center for Computer Science and Control of the Russian Academy of Sciences)
OptimizationComputational EfficiencyFlow-based ModelGraphBenchmark
🎯 What it does: A bulk search algorithm (Bulk Search) for anonymous multi-agent path planning based on flow networks is proposed, achieving faster path solving by compressing the search state.
Improved Bandits in Many-to-One Matching Markets with Incentive Compatibility
Fang Kong (Shanghai Jiao Tong University), Shuai Li (Shanghai Jiao Tong University)
Reinforcement Learning
🎯 What it does: This paper studies the Bandit learning problem in many-to-one matching markets and proposes the Adaptive Exploration-Deferred Acceptance (AETDA) algorithm to achieve the optimal stable regret upper bound for players, along with the Online Deferred Acceptance (ODA) algorithm under alternative preferences.
Improved Graph Contrastive Learning for Short Text Classification
Yonghao Liu (Jilin University), Renchu Guan (Jilin University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningText
🎯 What it does: This paper proposes a short text classification model named GIFT, which combines heterogeneous graph learning and contrastive learning, and introduces SVD dimensionality reduction to generate augmented views and seed K-means based weak label assignment to enhance representation learning.
Improved Metric Distortion via Threshold Approvals
Elliot Anshelevich (Rensselaer Polytechnic Institute), Alexandros A. Voudouris (University of Essex)
🎯 What it does: This paper proposes and analyzes a deterministic decision-making mechanism that combines threshold approval sets (α-TAS) with traditional ordinal information in metric spaces, aiming to reduce social costs and the metric distortion of maximum costs.
Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding
Yanmei Zou (Hunan University), Naveed Akhtar (University of Melbourne)
ClassificationObject DetectionSegmentationPoint Cloud
🎯 What it does: A multi-layer perceptron network called HPENet based on high-dimensional position encoding (HPE) is proposed for 3D classification, segmentation, and detection tasks of point clouds.
Improving Audio-Visual Segmentation with Bidirectional Generation
Dawei Hao (Bilibili Inc.), Yiran Zhong (NIO)
SegmentationTransformerContrastive LearningVideoMultimodalityBenchmarkAudio
🎯 What it does: By introducing a bidirectional generative framework and a visual-to-audio projection module, pixel-level segmentation of audible objects in videos is achieved.
Improving Automatic VQA Evaluation Using Large Language Models
Oscar Mañas (Mila), Aishwarya Agrawal (Mila)
TransformerLarge Language ModelPrompt EngineeringTextMultimodality
🎯 What it does: This paper proposes LAVE, an automatic evaluation metric for visual question answering based on instruction-tuned large language models.
Improving Cross-Modal Alignment with Synthetic Pairs for Text-Only Image Captioning
Zhiyue Liu (Guangxi University), Fanrong Ma (Guangxi University)
GenerationData SynthesisOptimizationTransformerDiffusion modelContrastive LearningImageTextMultimodality
🎯 What it does: A text-only trained image captioning model called SynTIC is proposed, which achieves cross-modal alignment by generating synthetic images and optimizing pseudo-image features in the CLIP space.
Improving Diffusion-Based Image Restoration with Error Contraction and Error Correction
Qiqi Bao (Tsinghua University), Wenming Yang (Tsinghua University)
RestorationSuper ResolutionConvolutional Neural NetworkDiffusion modelImageStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: This paper proposes the DiffECC method, which introduces CNN priors, restart sampling, and error correction strategies into the reverse sampling process of diffusion models to achieve high-quality image restoration.
Improving Distinguishability of Class for Graph Neural Networks
Dongxiao He (Tianjin University), Zhiyong Feng (Tianjin University)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes two metrics based on the Distinguishability of Class (D.C.), namely LDC and GDC, and utilizes them to guide graph neural networks in learning more distinguishable node representations, thereby enhancing node classification and clustering performance.
Improving Expressive Power of Spectral Graph Neural Networks with Eigenvalue Correction
Kangkang Lu (Beijing University of Posts and Telecommunications), Tat-Seng Chua (National University of Singapore)
Graph Neural NetworkGraph
🎯 What it does: The paper proposes an adaptive eigenvalue correction strategy to reduce the number of duplicate eigenvalues in the normalized Laplacian matrix, thereby enhancing the expressive power and fitting performance of polynomial spectral graph neural networks.
Improving Factual Error Correction by Learning to Inject Factual Errors
Xingwei He (University of Hong Kong), Siu Ming Yiu (University of Hong Kong)
GenerationData SynthesisTransformerLarge Language ModelText
🎯 What it does: A method named LIFE for unsupervised factual error correction is proposed, which automatically generates error-correct text pairs and performs corrections using a 'mask-then-corrupt-then-correct' process.
Improving GNN Calibration with Discriminative Ability: Insights and Strategies
Yujie Fang (Beijing Institute of Technology), Mingzhong Wang (University of the Sunshine Coast)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes a post-hoc calibration method based on discriminative capability, DC(GNN), which simultaneously reduces ECE and enhances AUROC by integrating multiple discriminative signals.
Improving Knowledge Extraction from LLMs for Task Learning through Agent Analysis
James R. Kirk (Integrated Cognition), John E. Laird (Integrated Cognition)
Robotic IntelligenceTransformerLarge Language ModelAgentic AIPrompt EngineeringText
🎯 What it does: A method called STARS is proposed, allowing embedded agents to utilize large language models (LLM) to obtain executable task goal descriptions in a single learning instance, and to evaluate, repair, and select feasible answers through the agent's own reasoning mechanism.
Improving Neural Network Generalization on Data-Limited Regression with Doubly-Robust Boosting
Hao Wang (Zhejiang University)
OptimizationConvolutional Neural NetworkRecurrent Neural NetworkGraph Neural NetworkTime Series
🎯 What it does: A late-stage optimization framework called DRBoost is proposed, which combines statistical learners and zero-order optimizers to recalibrate neural networks trained with SGD, enhancing their generalization performance.
Improving Open Set Recognition via Visual Prompts Distilled from Common-Sense Knowledge
Seongyeop Kim (Korea Advanced Institute of Science and Technology), Yong Man Ro (Korea Advanced Institute of Science and Technology)
RecognitionKnowledge DistillationConvolutional Neural NetworkLarge Language ModelContrastive LearningImage
🎯 What it does: Utilizing large language models to extract and distill general common sense knowledge, generating visual prompts corresponding to each known category, and aligning them with visual model features to enhance the unknown category detection capability of visual models in open set recognition (OSR).
Improving Open-Domain Dialogue Response Generation with Multi-Source Multilingual Commonsense Knowledge
Sixing Wu (Yunnan University), Wei Zhou (Yunnan University)
GenerationTransformerSupervised Fine-TuningText
🎯 What it does: Proposes a multi-source multi-lingual knowledge-driven dialogue generation task (MMKRG) and its dataset MMK-DailyDialog, and introduces the MMK-BART model to address cross-language conflicts and repetition issues.
Improving Panoptic Narrative Grounding by Harnessing Semantic Relationships and Visual Confirmation
Tianyu Guo (Xiamen University), Xiaoshuai Sun (Xiamen University)
Object DetectionSegmentationConvolutional Neural NetworkTransformerImageMultimodality
🎯 What it does: A one-stage Panoptic Narrative Grounding method, XPNG, is proposed, which utilizes semantic and visual relationships to achieve more accurate instance and region localization.
Improving PTM Site Prediction by Coupling of Multi-Granularity Structure and Multi-Scale Sequence Representation
Zhengyi Li (Huazhong Agricultural University), Wen Zhang (Huazhong Agricultural University)
Protein Structure PredictionConvolutional Neural NetworkRecurrent Neural NetworkTransformerContrastive LearningBiomedical Data
🎯 What it does: A PTM site prediction method called PTM-CMGMS is proposed, which combines multi-granularity structural information (atom, amino acid, whole protein) with multi-scale sequence representation;
Improving Robustness for Joint Optimization of Camera Pose and Decomposed Low-Rank Tensorial Radiance Fields
Bo-Yu Chen, Yu-Lun Liu (National Yang Ming Chiao Tung University)
Pose EstimationOptimizationNeural Radiance FieldImage
🎯 What it does: A neural field representation based on decomposed low-rank tensors is proposed, which jointly optimizes camera pose and scene geometry under the condition of having only 2D image supervision.
Improving the Adversarial Transferability of Vision Transformers with Virtual Dense Connection
Jianping Zhang (Chinese University of Hong Kong), Michael R. Lyu (Chinese University of Hong Kong)
Adversarial AttackTransformerImage
🎯 What it does: By adding a Virtual Dense Connection (VDC) scheme to the Vision Transformer (ViT), the transferability of adversarial samples across models and structures is enhanced.
Improving the Robustness of Knowledge-Grounded Dialogue via Contrastive Learning
Jiaan Wang (Soochow University), An Liu (Hong Kong Polytechnic University)
TransformerContrastive LearningText
🎯 What it does: This paper proposes an entity-driven contrastive learning framework called EnCo, aimed at enhancing the robustness of knowledge-driven dialogue systems when faced with real noise (such as typos, incomplete or incorrect entities).
Improving Transferability for Cross-Domain Trajectory Prediction via Neural Stochastic Differential Equation
Daehee Park (KAIST), Kuk-Jin Yoon (KAIST)
Domain AdaptationAutonomous DrivingRecurrent Neural NetworkTime SeriesSequentialStochastic Differential Equation
🎯 What it does: A cross-domain trajectory prediction framework based on Continuous Stochastic Differential Equations (NSDE) is proposed, which can simultaneously handle different time step configurations and trajectory noise generated by the data collection process, and achieve adaptive robustness to trajectory errors through dataset-specific diffusion networks.
In-Hand 3D Object Reconstruction from a Monocular RGB Video
Shijian Jiang (Zhejiang University), Jiming Chen (OPPO US Research Center)
RestorationObject DetectionSegmentationPose EstimationConvolutional Neural NetworkNeural Radiance FieldVideo
🎯 What it does: 3D reconstruction of handheld objects using monocular RGB videos, with complete geometry restored in occluded areas through implicit surface models.
Incomplete Contrastive Multi-View Clustering with High-Confidence Guiding
Guoqing Chao (Harbin Institute of Technology), Dianhui Chu (Harbin Institute of Technology)
OptimizationRepresentation LearningGraph Neural NetworkContrastive LearningImage
🎯 What it does: An end-to-end missing multi-view clustering method called ICMVC is proposed, which can simultaneously handle missing value processing, representation learning, and clustering assignment.
Inconsistency-Based Data-Centric Active Open-Set Annotation
Ruiyu Mao (University of Texas at Dallas), Yunhui Guo (University of Texas at Dallas)
ClassificationComputational EfficiencyData-Centric LearningConvolutional Neural NetworkLarge Language ModelContrastive LearningImage
🎯 What it does: This paper proposes a data-driven active open-set annotation method called NEAT, which utilizes the CLIP pre-trained model to extract features, performs known class detection based on kNN clustering of known classes, and uses model predictions and local feature distribution inconsistencies as active sampling criteria to actively select and annotate effective samples from an unlabeled data pool containing unknown classes.
Incremental Quasi-Newton Methods with Faster Superlinear Convergence Rates
Zhuanghua Liu (National University of Singapore), Bryan Kian Hsiang Low (National University of Singapore)
OptimizationTabular
🎯 What it does: Two incremental quasi-Newton methods are proposed: LISR-1 (based on SR1 update) and its extension LISRk (based on SRk update), achieving local superlinear convergence for large-scale strongly convex finite problems without relying on the condition number.
Independence of Irrelevant Alternatives under the Lens of Pairwise Distortion
Théo Delemazure (CNRS, Paris Dauphine University, PSL), Grzegorz Pierczyński (University of Warsaw)
🎯 What it does: This paper proposes and quantitatively evaluates a tool for assessing the impact of IIA on social welfare—pairwise distortion—and compares the distortions of various voting rules under average and worst-case scenarios.
Independency Adversarial Learning for Cross-Modal Sound Separation
Zhenkai Lin (University of Electronic Science and Technology of China), Yang Yang (University of Electronic Science and Technology of China)
RecognitionGenerative Adversarial NetworkContrastive LearningMultimodalityAudio
🎯 What it does: This paper proposes an unsupervised audio separation framework called IAL-CMS, based on independence adversarial learning and cross-modal speech separation.
Inducing Clusters Deep Kernel Gaussian Process for Longitudinal Data
Junjie Liang (Pennsylvania State University), Vasant Honavar (Pennsylvania State University)
Gaussian SplattingTime SeriesSequentialBiomedical DataAlzheimer's Disease
🎯 What it does: This paper proposes a deep kernel Gaussian process model named ICDKGP, designed to handle irregular, sparse, and potentially abrupt longitudinal data, by modeling continuity and abrupt changes through the combination of zero-mean GP and a deterministic mean function.
Inducing Point Operator Transformer: A Flexible and Scalable Architecture for Solving PDEs
Seungjun Lee (Alsemy), TaeiL Oh
TransformerMeshPhysics Related
🎯 What it does: This paper proposes the Inducing Point Operator Transformer (IPOT), a PDE operator learning framework that can handle arbitrary input-output discretization and is scalable.
Inertial Algorithm with Dry Fraction and Convolutional Sparse Coding for 3D Localization with Light Field Microscopy
Xiaofan Wang (Yanshan University), Jinjia Wang (Yanshan University)
OptimizationComputational EfficiencyImageBiomedical Data
🎯 What it does: A Fast-IPGDF algorithm based on inertial proximal gradient damping and Nesterov acceleration is proposed to improve the convolutional sparse coding (CSC) framework for 3D neuron point source localization in light field microscopy (LFM).
Inference and Learning in Dynamic Decision Networks Using Knowledge Compilation
Gabriele Venturato (KU Leuven), Luc De Raedt (KU Leuven)
Reinforcement LearningTabular
🎯 What it does: A dynamic decision circuit (DDC) based on knowledge compilation and a corresponding value iteration algorithm mapl-cirup are proposed to perform Bellman updates in dynamic decision networks (DDN), completing MDP planning and offline reinforcement learning.
Influential Exemplar Replay for Incremental Learning in Recommender Systems
Xinni Zhang (Chinese University of Hong Kong), Irwin King (Chinese University of Hong Kong)
Recommendation SystemTabular
🎯 What it does: An incremental learning framework called INFER has been developed, which uses influence functions to select representative samples for experience replay, maintaining the recommendation model's memory of early knowledge while adapting to new trends.
Information Design for Congestion Games with Unknown Demand
Svenja M. Griesbach (Technische Universität Berlin), Tim Koglin (Goethe Universität Frankfurt)
OptimizationGraph
🎯 What it does: This paper studies how information design can influence players' equilibrium behavior through public signals in non-autonomous network congestion games with unknown demand, and proposes an algorithm for optimizing signal schemes to minimize total expected costs. For single-source single-target networks with linear costs, it provides a Fully Polynomial-Time Approximation Scheme (FPTAS) for two states, a graph structure determination for complete information disclosure optimality, and a linear programming (LP) solving framework based on support sets; experiments are conducted on real traffic network instances for validation.
INFORMEDQX: Informed Conflict Detection for Over-Constrained Problems
Viet-Man Le (Graz University of Technology), Mathias Uta (Siemens Energy AG)
OptimizationTabular
🎯 What it does: This paper proposes the INFORMEDQX algorithm, which improves QUICKXPLAIN to utilize historical conflict knowledge for conflict detection.
Input Margins Can Predict Generalization Too
Coenraad Mouton (North-West University), Marelie H Davel (North-West University)
ClassificationImageBenchmark
🎯 What it does: A 'constrained margin' metric based on principal component subspace constraints in the input space is proposed to predict the generalization performance of deep networks.
Inspecting Prediction Confidence for Detecting Black-Box Backdoor Attacks
Tong Wang (Nanjing University), Ting Wang (Stony Brook University)
Anomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: A method named DTINSPECTOR for backdoor detection and elimination is proposed, which detects the impact of black-box backdoor attacks on the prediction confidence of trained models.
Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning
Yang Jiao (Fudan University), Yu-Gang Jiang (Fudan University)
Object DetectionDepth EstimationAutonomous DrivingPoint Cloud
🎯 What it does: In multi-camera 3D detection, an instance-aware depth estimation is introduced, proposing the IA-BEV framework;
Instance-Conditional Timescales of Decay for Non-Stationary Learning
Nishant Jain, Pradeep Shenoy (Google Research India)
Reinforcement LearningImage
🎯 What it does: A method for instance-conditioned time scale decay for slow concept drift (MUSCATEL) is proposed, which allocates weights to training samples using a mixed exponential decay function and instance-specific scoring networks, thereby enhancing the forward transfer performance of batch learning in non-stationary environments.
InstructDoc: A Dataset for Zero-Shot Generalization of Visual Document Understanding with Instructions
Ryota Tanaka (NTT Corporation), Jun Suzuki (Tohoku University)
TransformerLarge Language ModelPrompt EngineeringImageText
🎯 What it does: This paper presents the InstructDoc dataset and the InstructDr model, designed to complete various visual document understanding tasks through natural language instructions.
Instrumental Variable Estimation for Causal Inference in Longitudinal Data with Time-Dependent Latent Confounders
Debo Cheng (University of South Australia), Thuc Duy Le (University of South Australia)
Recurrent Neural NetworkTime SeriesSequential
🎯 What it does: A time-dependent variable factor model (TIFM) based on LSTM is proposed, utilizing implicit instrumental variables to infer causal effects in long sequences.
Integer Is Enough: When Vertical Federated Learning Meets Rounding
Pengyu Qiu (Zhejiang University), Shouling Ji (Zhejiang University)
Federated LearningSafty and PrivacyComputational EfficiencyAdversarial AttackImage
🎯 What it does: A 'rounding layer' is proposed to convert intermediate results from floating-point numbers to integers in Vertical Federated Learning (VFL), reducing computational and communication overhead while enhancing privacy and robustness.
Integrated Decision Gradients: Compute Your Attributions Where the Model Makes Its Decision
Chase Walker (University of Central Florida), Rickard Ewetz (Lockheed Martin)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: A new path integral attribution method is proposed - Integrated Decision Gradients (IDG), which addresses the saturation problem of Integrated Gradients by weighting the gradients according to the output logit change rate.
Intelligent Calibration for Bias Reduction in Sentiment Corpora Annotation Process
Idan Toker (Bar-Ilan University), Jonathan Schler (Holon Institute of Technology)
Text
🎯 What it does: A calibration set-based sentiment labeling method is proposed to reduce anchoring bias during the serialization labeling process.
Intentional Evolutionary Learning for Untrimmed Videos with Long Tail Distribution
Yuxi Zhou (Tsinghua University), Shengyong Chen (Tianjin University of Technology)
ClassificationRecognitionRecurrent Neural NetworkVideo
🎯 What it does: This study investigates predicting human intentions in untrimmed videos and proposes the ICCA loss based on instance confidence and category accuracy, as well as an intention evolution learning method.
Interactive Hyperparameter Optimization in Multi-Objective Problems via Preference Learning
Joseph Giovanelli (Alma Mater Studiorum University of Bologna), Marius Lindauer (Institute of Artificial Intelligence L3S Research Center Leibniz University Hannover)
OptimizationHyperparameter SearchTabular
🎯 What it does: An interactive multi-objective machine learning hyperparameter optimization method is proposed, which allows users to compare different Pareto front binary examples to learn a personalized quality metric, and then uses this metric as a loss function to perform hyperparameter search in SMAC.
Interactive Visual Task Learning for Robots
Weiwei Gu (Arizona State University), Nakul Gopalan (Arizona State University)
Robotic IntelligenceGraph Neural NetworkVision Language ModelImageText
🎯 What it does: Through a human-robot interaction demonstration and language description, the robot learns new visual concepts and structures them into scene graphs, enabling zero-shot generation and completion of new tasks after a single demonstration.
Interpretable3D: An Ad-Hoc Interpretable Classifier for 3D Point Clouds
Tuo Feng (University of Technology Sydney), Yi Yang (Zhejiang University)
ClassificationExplainability and InterpretabilityPoint Cloud
🎯 What it does: An interpretable classifier for 3D point clouds, Interpretable3D, is proposed.
InterpretARA: Enhancing Hybrid Automatic Readability Assessment with Linguistic Feature Interpreter and Contrastive Learning
Jinshan Zeng (Jiangxi Normal University), Qing Huang (Jiangxi University of Science and Technology)
ClassificationExplainability and InterpretabilityRepresentation LearningTransformerContrastive LearningText
🎯 What it does: Proposes the InterpretARA mixed readability assessment model, which combines a language feature interpreter with contrastive learning to extract and integrate deep representations at the document and paragraph levels.
Intra- and Inter-group Optimal Transport for User-Oriented Fairness in Recommender Systems
Zhongxuan Han (Zhejiang University), Jianwei Yin (Zhejiang University)
Recommendation SystemTabular
🎯 What it does: This paper proposes the II-GOOT framework for the training phase and the ξ-UOF metric for the evaluation phase to address the performance gap between user groups (active and inactive) in recommendation systems (User-Oriented Fairness, UOF). The aim is to alleviate training bias caused by data sparsity and achieve fairer recommendation outcomes.
Intrinsic Action Tendency Consistency for Cooperative Multi-Agent Reinforcement Learning
Junkai Zhang (Institute of Automation, Chinese Academy of Sciences), Jian Cheng (Institute of Automation, Chinese Academy of Sciences)
Reinforcement LearningSequential
🎯 What it does: A new internal reward mechanism based on action tendency consistency is proposed and combined with the CTDE framework to enhance the training efficiency and performance of cooperative multi-agent reinforcement learning.
Intrinsic Phase-Preserving Networks for Depth Super Resolution
Xuanhong Chen (Shanghai Jiao Tong University), Bingbing Ni (Shanghai Jiao Tong University)
RestorationDepth EstimationSuper ResolutionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a phase-preserving deep depth map super-resolution network (IPPNet), which designs a phase-preserving filtering module (PPFM) through phase analysis to achieve depth-guided RGB noise filtering and phase alignment, thereby enhancing the super-resolution performance of depth maps.
Invariant Random Forest: Tree-Based Model Solution for OOD Generalization
Yufan Liao (Renmin University of China), Xing Yan (City University of Hong Kong)
ClassificationDomain AdaptationOptimizationExplainability and InterpretabilityTabularTime SeriesFinance Related
🎯 What it does: This paper proposes a decision tree and random forest model based on theoretical invariance (Invariant Decision Tree and Invariant Random Forest). By adding a cross-environment invariance penalty term to the tree splitting criterion, it suppresses the use of features that change with the environment, enhancing out-of-distribution (OOD) generalization ability.
Inverse Weight-Balancing for Deep Long-Tailed Learning
Wenqi Dang (XiDian University), Guangming Shi (Peng Cheng Laboratory)
ClassificationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: Proposes an Inverse Weight-Balancing (IWB) two-stage training framework, first training the network with cross-entropy, then freezing the encoder to fine-tune the classifier based on inverse distribution KL loss.
Investigating the Effectiveness of Task-Agnostic Prefix Prompt for Instruction Following
Seonghyeon Ye (Korea Advanced Institute of Science and Technology), Minjoon Seo (LG AI Research)
ClassificationGenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This study investigates a task-agnostic prefix prompt (TAPP) that is directly concatenated to the input during inference, significantly enhancing the instruction-following capability of large language models.
IOFM: Using the Interpolation Technique on the Over-Fitted Models to Identify Clean-Annotated Samples
Dongha Kim (Sungshin Women's University), Yongdai Kim (Seoul National University)
ClassificationData-Centric LearningConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an interpolation technique using overfitting models (IOFM) to identify clean labels in training data.
IPRemover: A Generative Model Inversion Attack against Deep Neural Network Fingerprinting and Watermarking
Wei Zong (University of Wollongong), Seyit Camtepe (CSIRO Data61)
Data SynthesisKnowledge DistillationAdversarial AttackGenerative Adversarial NetworkImage
🎯 What it does: A data-independent IP removal attack called IPRemover is proposed, which utilizes model inversion to generate training data and evades DNN fingerprint and watermark detection through virtual ensemble knowledge distillation.
IRPruneDet: Efficient Infrared Small Target Detection via Wavelet Structure-Regularized Soft Channel Pruning
Mingjin Zhang (Xidian University), Jing Zhang (University of Sydney)
Object DetectionComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a soft channel pruning method called IRPruneDet based on wavelet structural regularization, aimed at achieving lightweight and efficient infrared small target detection models.
Is a Large Language Model a Good Annotator for Event Extraction?
Ruirui Chen (Institute of High Performance Computing), Dongkyu Choi (Institute of High Performance Computing)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Using large language models (LLMs) as expert annotators to generate additional annotated samples that conform to the original data distribution for the event extraction task, thereby alleviating data scarcity and long-tail distribution issues, and improving model performance.
IS-DARTS: Stabilizing DARTS through Precise Measurement on Candidate Importance
Hongyi He (Xi'an Jiaotong University), Nanning Zheng (Xi'an Jiaotong University)
OptimizationNeural Architecture SearchImage
🎯 What it does: The IS-DARTS method is proposed, which evaluates the importance of candidate operations by measuring their Fisher information and combines a multi-step shrinking strategy to improve and stabilize DARTS.
ISP-Teacher:Image Signal Process with Disentanglement Regularization for Unsupervised Domain Adaptive Dark Object Detection
Yin Zhang (Harbin Institute of Technology), Mingli Ding (Harbin Institute of Technology)
Object DetectionDomain AdaptationAutonomous DrivingKnowledge DistillationImage
🎯 What it does: This paper proposes ISP-Teacher, which utilizes self-supervised ISP degradation learning and a discretization regularization Teacher-Student framework to achieve target detection in unlabeled low-light environments.
IT3D: Improved Text-to-3D Generation with Explicit View Synthesis
Yiwen Chen (Nanyang Technological University), Guosheng Lin (Tencent PCG)
GenerationData SynthesisDiffusion modelGenerative Adversarial NetworkImageMesh
🎯 What it does: Based on existing rough 3D models, high-quality multi-view renderings are first generated using an image-to-image diffusion model (ControlNet + Stable Diffusion), and then the 3D model is refined and edited using a dual training strategy with a discriminator and DiffusionGAN.
Iterative Regularization with k-support Norm: An Important Complement to Sparse Recovery
William de Vazelhes (Mohamed Bin Zayed University of Artificial Intelligence), Bin Gu (Nanjing University)
OptimizationTabular
🎯 What it does: An iterative regularization algorithm IRKSN based on the k-support norm is proposed to achieve fast recovery of sparse vectors.
Iterative Token Evaluation and Refinement for Real-World Super-resolution
Chaofeng Chen (Nanyang Technological University), Weisi Lin (Nanyang Technological University)
RestorationSuper ResolutionTransformerDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: An iterative label evaluation and refinement (ITER) framework based on a discrete diffusion model is proposed for super-resolution of real scene images. It first maps low-quality images to a high-quality label space through a dedicated denoising network, and then gradually refines textures using a discrete diffusion model with an evaluation module.
IVP-VAE: Modeling EHR Time Series with Initial Value Problem Solvers
Jingge Xiao (L3S Research Center, Leibniz Universitat Hannover), Sandipan Sikdar (Indian Institute of Technology Kharagpur)
ClassificationRepresentation LearningRecurrent Neural NetworkAuto EncoderTime SeriesBiomedical DataElectronic Health RecordsOrdinary Differential Equation
🎯 What it does: A model called IVP-VAE based on continuous-time variational autoencoders has been designed for modeling, predicting, and classifying unordered and sparse electronic health record (EHR) time series.
Joint Demosaicing and Denoising for Spike Camera
Yanchen Dong (Peking University), Tiejun Huang (Peking University)
RestorationConvolutional Neural NetworkImageVideo
🎯 What it does: A joint denoising and demosaicing (JDD) network is proposed for spike cameras with color filter arrays, capable of recovering high-quality color images from the binarized Bayer-pattern spike stream.
Joint Learning Neuronal Skeleton and Brain Circuit Topology with Permutation Invariant Encoders for Neuron Classification
Minghui Liao (Wuhan University), Bo Du (Wuhan University)
ClassificationGraph Neural NetworkGraph
🎯 What it does: A neuron classification framework called NeuNet was designed and implemented, which simultaneously utilizes neuron skeleton morphology and brain circuit topology information, and two whole-brain reconstruction datasets were made public.
Jointly Improving the Sample and Communication Complexities in Decentralized Stochastic Minimax Optimization
Xuan Zhang (Pennsylvania State University), Yangyang Xu (Rensselaer Polytechnic Institute)
OptimizationTabular
🎯 What it does: A single-cycle distributed gradient descent ascent algorithm DGDA-VR is proposed to solve non-convex-strongly concave minimax problems with multiple agents, distributed data, and access only to unbiased stochastic gradients.
Jointly Modeling Spatio-Temporal Features of Tactile Signals for Action Classification
Jimmy Lin (Institute for AI Industry Research), Yang Liu (Institute for AI Industry Research)
ClassificationTransformerTime Series
🎯 What it does: A Transformer model that can simultaneously capture spatial and temporal features of tactile signals (STAT) is proposed, enhancing action classification performance through spatial embedding, temporal embedding, and a temporal pre-training task.